| series | collection frequency | oldest observation | newest observation |
|---|---|---|---|
| 30-year fixed mortgage rate | weekly | April 2, 1971 | January 8, 2026 |
| consumer price index (cpi) | monthly | January 1, 1947 | November 1, 2025 |
| median home price | quarterly | January 15, 1953 | April 1, 2025 |
| median household income | annually | January 1, 1947 | January 1, 2024 |
how expensive is housing in the US?
I am actively updating this post. Content will change as I work my way through this article.
tl;dr
Things are bad, but not as bad as you might think.
At a national scale,
- From February 1953 to January 2026, the price-to-income ratio of a home has increased by 2.867 percentage points (pp) per year on average. As of January 2026, the ratio stands at 4.511, placing it in the 73rd percentile of all observations in this period.
- From April 1971 to January 2026, the mortgage debt-to-income ratio has decreased by -0.138 pp per year on average. As of January 2026, the ratio stands at 0.262, placing it in the 46th percentile of all observations in this period.
1 Meaning the typical house costs the typical household 4.51 \(\times\) their gross annual income
2 Meaning the typical household would spend 0.26 \(\times\) their monthly income on the typical mortgage. This calculation assumes only interest and principal.
intro
Much has been written recently on housing affordability, with much of it being negative.3 Aside from the tendency of news organizations to produce negative content, I think what is lost in these reports is historical context: where were we, and where are we now? Even when the historical context is present, metrics may be used that, if not interpreted carefully, can be misleading.
For instance, longtermtrends.com gives a chart that shows the ratio of the S&P/Case-Shiller Home Price Index to the median household income. This index is used a lot since it’s high-quality and has data that dates back to 1890. However, the problem here is that the Case-Shiller index is a weighted average. The distribution of home values tend to be right skewed, so an average will be greater than the median (the typical house). So, dividing the average by the median produces a price-to-income ratio that is a bit more pessimistic. The typical person is buying the typical house, not the average house.
In order to provide some clarity on the question of housing affordability, I’m going to look at two metrics:
- The ratio of the typical cost of a house to the typical household income. In this article, I’ll refer to it as the price-to-income ratio.
- The proportion of a person’s monthly income that would be dedicated to their monthly mortgage. For my analysis, this will be limited to only interest and principal4, and will assume a 20% downpayment, so a loan on 80% the value of a house. I’ll refer to it as the debt-to-income ratio in the article.
4 I’m interested in having a better picture of this by incorporating property taxes and home insurance costs if someone could get me the data…
analysis
I’m planning to examine this at three different geographic scales:
- national
- state
- county
national scale
the data
The data I’m using is mostly from the FRED. I’ve supplemented some of these series with data extending further back to get a more complete picture. Below are the data series and where I got the supplemental data:
- CPIAUCSL: Consumer Price Index for All Urban Consumers: All Items in U.S. City Average
- MEHOINUSA646N: Median Household Income in the United States
- Years 1947-1965 are from p.877 of a US Census Bureau publication.
- Years 1967-1983 are also from the US Census Bureau
- MORTGAGE30US: 30-Year Fixed Rate Mortgage Average5 in the United States
- MSPUS: Median Sales Price of Houses Sold for the United States
5 Unfortunately, there is no high-frequency median equivalent to this data that I know of. Fortunately, there are some restrictive filters (e.g. 20% down, excellent credit, excluding jumbo loans) on which data to use that removes large outliers making the mean more centered. See here for more details
Let’s take a look at the data below:
There are a few things that need to be handled before analysis:
- The data is captured at different frequencies.
- The data starts at different times.
- Not all the data is recent.7
7 The 30-year fixed mortgage rate is the most egregious offender. The reason for this, is that Freddie Mac began surveying lenders on mortgage rates in April 1971, making this the earliest date for which consistent, nationwide data is available.
To account for (1), I put everything in a monthly frequency. Cubic splines were used for quarterly and annual data, while weekly data was averaged. For (2), I used the maximum of the minimum date available (e.g. price-to-income ratio begins at 1953). For (3), I first smoothed each series using LOESS then modeled each series using ARIMA to produce estimates for the most recent date available (January 2026). These estimates should be taken with a grain of salt, though I include \(3 \sigma\) prediction intervals to capture the uncertainty.
These smoothed indicators are used to calculate the metrics in the following sections.
price-to-income ratio
We can use nominal dollars to calculate the price to income ratio. I’ll include a linear model to help us understand the trend over time.
debt-to-income ratio
scratch-work

What is the linear trend in the inflation adjusted median sales price of a home? Let’s exclude the forecast data:
Call:
lm(formula = ia2025_price ~ date, data = tb_msia %>% filter(combined_sd ==
0))
Residuals:
Min 1Q Median 3Q Max
-45985 -23085 -1404 18327 71859
Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 2.263e+05 1.200e+03 188.54 <2e-16 ***
date 9.272e+00 1.160e-01 79.92 <2e-16 ***
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Residual standard error: 26020 on 865 degrees of freedom
Multiple R-squared: 0.8807, Adjusted R-squared: 0.8806
F-statistic: 6388 on 1 and 865 DF, p-value: < 2.2e-16

From the above model, we find that real median sales price (2025 dollars) of a home has increased by $3,384.26 a year on average from February, 1953 to April, 2025.
For the monthly mortgage payment (30-year mortgage), we’ll assume a 20% down payment so the mortgage will 80% of the home value:

Now, let’s get the percentage of a person’s gross income that would go to their mortgage:


Lastly, let’s look at the total cost of a house divided by the annual income:

Let’s build a model of the price/income ratio:
Call:
lm(formula = price_income_ratio ~ date, data = tb_final %>% filter(combined_sd ==
0))
Residuals:
Min 1Q Median 3Q Max
-0.7797 -0.3001 -0.1132 0.2582 1.6341
Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 3.415e+00 2.055e-02 166.16 <2e-16 ***
date 7.961e-05 1.987e-06 40.07 <2e-16 ***
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Residual standard error: 0.4457 on 865 degrees of freedom
Multiple R-squared: 0.6499, Adjusted R-squared: 0.6495
F-statistic: 1606 on 1 and 865 DF, p-value: < 2.2e-16

From the above model, we find that the ratio of the typical cost of a house to the typical household income has increased by 0.03 a year on average from February, 1953 to April, 2025.